The conditional increase in probability for "participated in combat sports or something non-surprisingly adjacent" based on "has cauliflower ear" may not be anywhere near as large as you think. Even just among the people reading this post, you've got someone with a cauliflower ear (me) who got it by something completely unrelated (the ear freezing solid). How can you tell if the sample you used to derive your explanation is biased? From the post, it sounds like you're into MMA more than is the average for the population. I would expect many things in your life therefore correlate with MMA just based on exposure, maybe completely spuriously, maybe just more strongly than is true for the population in general. If you're excluding entire segments of the population from your sample, and I think you necessarily must do so in any non-trivial application of this technique, I don't know how you could have confidence in the resulting explanation. At least not outside of very homogeneous populations.
It is persistent (or at least has persisted for about 20 years), is just the one ear (because that was what was exposed), and could perhaps be differentiated from blunt trauma cases in that the swelling only affects the helix and the directionality is different. How to explain the last part? If you imagine an axis going down your ear canal, with blunt trauma the swelling is generally parallel to that axis (the ear is made thicker) while with my ear it's along a radius perpendicular to that axis (the ear is made taller).